“Genetic”, or “evolutionary” algorithms are inspired from the evolution of species and from natural selection to bring solutions to optimization problems (such as, in the case of the districting of a territory: “how can I create N sectors with a balanced number both of clients and of prospects, all the while keeping a geographic coherence”). They are connected to Machine Learning, since they are “learning” algorithms, which optimize their results in the course of their successive executions.

To better understand and illustrate genetic algorithms, let us review succinctly Darwin’s teachings:

• The evolution of species stems from random genetic mutations (example: a different coat color compared to the rest of the species).
• Sometimes, these random genetic mutations give a competitive advantage to individuals (example: a different coat color allowing them to better blend with their environment) compared to their peers.
• These competitive advantages give them better chances of survival (examples: by hiding better, they evade predators more easily, become better hunters, etc.) and allow them to reproduce more: this is natural selection.
• In this way, these unusual, but “more competitive” genetic mutations tend towards generalizing to the whole of the species as generations go by.

Genetic algorithms draw from this idea to:

• Generate a “population” of initial solutions (for example, a thousand different districtings for a territory), which are more or less random or subject to constraints identified beforehand.
• Subject these initial solutions to tests (for example, to objectives regarding the balance in the number of clients or prospects, the maximum distance to be covered, the compacity, the geographic cohesion, etc.).
• Only retain the best solutions (those which are closest to the specified objectives and constraints).
• Generate a new “generation” of solutions by cross-referencing/applying mutations to the best solutions of the prior generation (creating new versions of the districting of a territory, but closer to the objective than the prior generation).
• Repeat this process of selection/mutation for several generations, until it reaches a set of “optimum” solutions.

Genetic algorithms are a method used in the field of Artificial Intelligence (AI). Applied to the districting of a territory, they make it possible to obtain optimized solutions even with complex constraints.